What is AI PC?

AI is everywhere. AI companies like OpenAI and Anthropic went from unknown to household names in just a few short years. Big Tech companies like Google, Meta, Amazon, and Microsoft continue to roll out their AI models to every product, from search to chatbot assistants and more. And, naturally, companies that make computer chips and GPUs like Nvidia make billions as they make the hardware that powers it all.

But what if consumers don’t want to send their data to Big Tech companies or fork over hundreds, if not thousands, of dollars a month to Big AI token companies? While computer hardware companies like Nvidia may be struggling to keep up with demand, everyday users can still find everything they need to build their own machine – an AI PC – and run AI models locally.

But what exactly is an AI PC? What do you need if you want to build your own, or what should you look for when buying a pre-made AI PC? And how does it differ from other powerful computers built for specific purposes, such as gaming PCs?

To answer these questions, Mashable spoke to an expert who builds PCs for a living, Quoted Tech founder and CEO Kevin Jia. Established in 2023, Quoted from Tech is a Canadian company that builds custom PCs for thousands of customers. Jia and company also provided Mashable with access to one of their pre-built models, the Quoted from One Pro Plusan entry-level pre-built computer designed for AI, so we can explore what makes an AI PC an AI PC.

What is AI PC?

The AI ​​PC is exactly what it says it is: A computer built to handle artificial intelligence, machine learning, and large-scale linguistic models (LLMs).

Technically, every PC can run AI these days, just as one can play games or edit videos on any computer, no gaming PC or video editing workstation is needed. Just as a gaming PC delivers high performance and a great gaming experience, an AI PC is specifically designed to run AI models.

As Jia pointed out, however, an AI PC is not meant to be a single-purpose machine, as most of the information you can get from an AI PC can be found in a gaming machine or a PC built for video processing, too.

“Users don’t have to choose between an ‘AI’ rig and a ‘gaming’ rig,” Jia explained. “An AI-centric computer doesn’t have to be a single-purpose machine. It’s a workstation that’s fully capable of AI work, but also designed to comfortably handle gaming and other professional tasks in a single system.”

Can I use my laptop for AI?

Yes, you can run ChatGPT, Claude, Copilot, and Gemini all on your $300 laptop. Cloud models can run on any machine.

However, if you’re trying to use local AI processes, you’ll need a PC built for the job.

“AI needs a lot of GPU processing power, and you need a lot of VRAM, and you need a lot of memory, and you need a decent CPU, and you need to be able to cool all of that in a decent tower,” Jia said. “We’ve had customers come to us and say, ‘Hey, I bought this AI laptop, and it doesn’t work, it just gets too hot, and it’s slow.’

Quoted from One Pro Plus
Credit: Mashable / Matt Binder

To get an idea of ​​what’s inside a consumer-grade AI PC, here are the specifications of the Quoted One Pro Plus we tested for this piece:

  • CPU: Intel Core i5-14600K (14 cores / 20 threads, up to 5.3 GHz)

  • GPU: NVIDIA RTX 5060 (8GB GDDR7)

  • Memory: 32GB DDR5 (2×16GB) Kingston Fury Beast @ 5600 MHz

  • Storage: 1TB Kingston KC3000 NVMe M.2 SSD (up to 7,000 MB/s read speed)

  • Motherboard: MSI PRO Z790-VC WiFi

  • CPU cooling: Thermalright Peerless Assassin 120 (white, dual-tower air cooler)

  • Power supply: MSI MAG A650BN 650W, 80+ Bronze (non-module)

  • Case: Fractal Design North (white with tempered glass side panel)

  • Operating System: Windows 11 Home

What is the main difference between a PC built for AI and other PC architectures?

Traditionally, when one thinks of a high-end PC, at least at the consumer level, they think of a tricked-out gaming PC.

While you can play on an AI PC and vice versa, custom PC builders like Quoted Tech know that there is a difference when building a PC for a specific purpose like AI.

One big difference? Emphasis on cooling.

“You can see there’s a ton of airflow,” Jia said. “That’s going to be a big non-negotiable for AI-powered systems.”

Fractal Design's North PC case ventilation

Fractal Design’s North ventilation
Credit: Fractal Design

After spending some time testing the Quoted One Pro Plus, “a ton of airflow” actually seems like an understatement. For its line of AI computers, Quoted Tech uses Fractal Design’s North PC case.

The first thing you’ll notice is that the PC tower is a lot bigger than you might expect. Yes, they could make it smaller. This is not a technical limitation. The PC case is deliberately large. The North is designed for maximum airflow to cool your PC.

“A lot of pre-built systems are built in a small way because it’s cheap,” Jia said. “We’ve had people come to us and say, ‘I’m using this in a pre-built game, and it just crashes or slows down, or doesn’t work.’ And we tell them they have the size of a large laptop. That program was not designed to be pushed like this.”

You can feel the breeze from the Quoted One Pro Plus when it’s running. It’s like night and day compared to my, very small gaming PC. My PC tower can heat my room if needed in the winter while I play a few minutes of Fortnite. Conversely, the airflow felt from the Quoted One Pro Plus can cool a small room in the summer.

“When you play, I’m sure most of us play for a few hours at a time,” Jia explained. “Many gaming systems, even though they’re good for airflow, may not be the best because they’re not designed to be used for 7 or 8 or 12 hours.”

“Like with AI, if you’re doing an LLM project or something like that, you might use this thing for 2 or 3 days.” he continued. “You need stability. You need coolness in the workplace. You need a working-class environment. The ones you can’t sell.”

There is also often a limit to how much processing power can go into an AI PC. For example, Quoted Tech has built workstations that can handle up to 8 GPUs at once.

“That would be a very bad gaming PC because most games won’t be able to use more than one GPU effectively,” Jie said. “If you say, ‘I’m going to buy a Blackwell card to play computer games,’ I’d say I think we need to rethink what you’re spending money on because you can probably get 99 percent of the performance. [in a gaming PC] one-third of the budget.”

Fractal Design's North PC case

Fractal Design’s North PC case
Credit: Fractal Design

Why AI PCs over Apple’s Silicon-powered Macs?

In November 2020, Apple released its first line of Silicon Macs with the new M1 MacBook Air, MacBook Pro, and Mac mini. These Macs offered consumers more powerful options than similarly priced laptops and desktop PCs.

Since Apple has packed a lot of processing power into these Silicon chips over the years, they are popular with users who want to run local AI processes such as an AI assistant. OpenClaw.

While these Macs are still good choices for certain use cases, Jia explains why PCs are still the top choice for AI.

“In general, most LLM and AI developers today are building with NVIDIA’s CUDA language. That’s actually a clear difference,” Jia said, referring to Nvidia’s proprietary Compute Unified Device Architecture, which requires Nvidia GPUs and thus cannot be run on Macs.

“Macs and Radeon-based systems just don’t have the same CUDA-driven ecosystem for developers to work within,” he continued. “For anyone building or seriously experimenting with AI models, that ecosystem functionality, determines what tools you can use, what workflows you can develop, and how easily you can scale your work.”

And, of course, PCs can be continuously upgraded with new hardware while consumers can upgrade the internals of their Macs.

To compare the two, I used a generative AI model of the images in both. Both completed the tasks based on the assigned instructions. This was not surprising. After all, both are designed to handle AI processing.

While the consumer-grade hardware I had wasn’t the best way to show the difference between PCs and Macs when it comes to AI processing, I took the opportunity to get my own Mac mini M4 Pro and Quoted Tech’s Quoted One Pro Plus. To compare the two, I used a generative AI model of the images in both. Both completed the tasks based on the assigned instructions. This was not surprising. After all, both are designed to handle AI processing.

However, the difference in the heat from the Mac mini and the cool air flowing through the Quoted One Pro Plus was hard to miss. It’s easy to see that the AI ​​PC’s form factor will give it an advantage over the Mac in terms of performance.

How much will an AI PC cost me?

Obviously, there is a wide range of prices, as AI PCs can be built for consumer use or large business use. Jia told me that Quoted Tech has built workstations that cost as much as six figures. But individual users who want to use local AI models obviously don’t need to spend anywhere near that amount.

Quoted Tech says that those who want to “get involved” in using local AI models can start with an entry-level AI, like the Quoted One Pro Plus, for between $1,500 and $2,500.

Articles
Artificial intelligence

Leave a Comment